COF-C03 domain - 10% of the exam

Data Collaboration

Data Collaboration is 10% of the COF-C03 exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

28
Practice questions
4
Objectives
7
Easy
21
Medium

Where people go wrong

  • Treating Fail-safe as self-service Time Travel

    The wrong answers reach for Fail-safe, a support case or a longer retention setting to undo a recent change. Time Travel is the self-service tool: AT or BEFORE reads an earlier state and UNDROP restores a dropped table while it is inside the retention window. Fail-safe is a fixed seven-day period for permanent tables, not user-queryable, and only Snowflake support can act on it.

    See it in a sample question
  • Confusing replication groups with failover groups

    The wrong answers treat the group types as interchangeable. Both a replication group and a failover group replicate a chosen set of databases and account objects to a target account in another region. A replication group gives read-only secondaries for reporting and cannot be promoted, and a failover group has no setting that turns promotion off. Only a failover group adds failover, so the secondary can be promoted to primary and serve writes during an outage.

    See it in a sample question
  • Expecting a share to write or fail over

    The wrong answers offer a data share as the way to move production to another account or to resume writes. A share gives read-only access to live data and carries no promotion mechanism. Moving the primary role is done with failover and failback on a failover group: refresh the original account from the current primary, then promote it back.

    See it in a sample question
  • Mixing up Marketplace listing types

    The wrong answers attach the wrong behaviour to a listing type. A public listing is discoverable by every Snowflake customer, so it restricts nothing. A private listing names the accounts that can see it, but privacy alone creates no charge. Only a paid listing lets the provider set pricing while Snowflake collects the fees. A direct share needs known consumer accounts, so it offers no open discovery.

    See it in a sample question
  • Confusing the application package with other sharing objects

    The wrong answers offer a secure share, a reader account, a scheduled task or the application object where provider logic must be packaged and run inside the consumer account. A share grants read access to data, a task runs only in its own account, and a reader account serves consumers with no Snowflake account. In the Native App Framework the provider builds an application package holding the manifest, setup script and versioned logic, and installation creates the application object in the consumer account.

    See it in a sample question

Objectives in this domain

Sample questions from this domain

Free sampleData Collaborationmedium

An analyst accidentally ran a bad UPDATE on a permanent table eight hours ago. The table has the default Time Travel retention and no clone or backup exists. The analyst wants to recover the pre-update values with the least effort. What should they do?

  • AOpen a support case to retrieve the previous values from Fail-safe, because Fail-safe is the standard self-service mechanism for reverting a recent erroneous update on a permanent table
  • BQuery the table using Time Travel with AT or BEFORE to read its state from just before the bad update, then restore those rows, since the change is within the retention window Correct
  • CRun UNDROP TABLE on the affected table, which rolls the table back to the version that existed before the most recent update statement was committed against it
  • DIncrease DATA_RETENTION_TIME_IN_DAYS on the table now, which retroactively extends the window backwards and recreates the pre-update version so it can be queried again
Use Time Travel AT or BEFORE within the retention window to recover prior table state after an erroneous DML statement. Time Travel lets you query and restore a table's earlier state within its retention period, which defaults to one day for a permanent table; because the bad update happened eight hours ago, the prior version is still retained and an AT or BEFORE query returns it for a self-service restore without involving Snowflake support.

Why A is wrong: Fail-safe is tempting as a recovery tool, but it is a Snowflake-managed last resort accessed only via support, not a self-service way to undo a recent change still in Time Travel.

Why B is correct: Default retention is one day, so an eight-hour-old change sits inside the Time Travel window and AT or BEFORE returns the prior version for a direct restore.

Why C is wrong: UNDROP only restores a dropped object; the table was updated, not dropped, so UNDROP does nothing here and cannot revert an UPDATE statement.

Why D is wrong: Raising retention only affects history kept from now on; it cannot recreate a version that the original window already covers or restore data retroactively.

Free sampleData Collaborationmedium

A platform team wants to keep a copy of several analytics databases in a second Snowflake account in another region purely for read scaling and reporting, and they are explicit that they do not need the ability to promote that second account to take over production during an outage. Which object should they create to group and replicate these databases together while matching this read-only intent?

  • AA share group that bundles the databases into a single Marketplace listing, so the reporting account subscribes once and receives ongoing read-only updates to all of the analytics databases together
  • BA failover group containing the analytics databases, then disabling its promotion setting, so the second account holds read-only copies for reporting while failover stays permanently switched off
  • CA standard view in the second account that selects across the analytics databases over an external connection, so the reporting account reads current data without copying it between the two regions
  • DA replication group that replicates the selected analytics databases together to the second account on a schedule, providing read-only secondaries for reporting without the ability to fail over and promote Correct
Distinguish a replication group, which replicates databases without failover, from a failover group, which adds the ability to promote a secondary to primary. Both replication groups and failover groups replicate a defined set of databases and account objects together to a target account. The key difference is that a replication group provides read-only secondaries without the ability to fail over, while a failover group adds promotion for disaster recovery. When the requirement is read scaling and reporting with no need to promote the second account, a replication group is the correct and best-matching object.

Why A is wrong: There is no share group object, and Marketplace listings publish shares for consumers rather than replicating databases into a specific second account, so this does not match the cross-account replication intent.

Why B is wrong: A failover group is tempting because it also replicates groups of objects, but its defining purpose is failover; there is no setting that converts it into a non-failover replication object, so it does not best match the stated read-only intent.

Why C is wrong: A standard view cannot reach across accounts or regions to read another account's databases, and it copies no data, so it cannot provide local replicated copies for read scaling in the second account.

Why D is correct: A replication group replicates a chosen set of databases and account objects as one unit to a target account but does not support failover, giving exactly the read-only reporting secondaries the team asked for without promotion.

Free sampleData Collaborationmedium

A company failed over to its secondary account in another region after a regional incident, and that secondary has been serving production writes for two days. The primary region is now healthy again and the company wants to return production operation to the original account with the least disruption. What is the correct way to move the primary role back to the original account?

  • ADrop the failover group in the original account and recreate it as a fresh secondary, then load a full copy of the data from the current primary before promoting it back to the primary role
  • BPerform a failback by refreshing the original account's secondary from the current primary and then promoting the original account's failover group back to primary, so production returns to the original region Correct
  • CCreate a secure data share from the current primary back to the original account so it can read the latest data, then switch client connections back to the original account once it is current
  • DSuspend replication entirely and manually export the changed rows from the current primary, then import them into the original account before pointing applications back to that region
Recognise that failback refreshes the original account as a secondary from the current primary and then promotes it back to the primary role. Failback reverses a failover. While the secondary served production, the original account became a secondary in the replication relationship. To return, you refresh that original secondary from the current primary so it catches up on changes made during the outage, then promote its failover group back to primary. This managed sequence avoids a full re-seed and keeps the data consistent, unlike dropping the group, sharing data, or hand-copying rows.

Why A is wrong: Dropping and recreating the group forces an unnecessary full re-seed of all data and discards replication state, which is far more disruptive than refreshing and promoting the existing secondary.

Why B is correct: Failback is the supported procedure: the original account, now a secondary, is refreshed from the current primary and then its failover group is promoted back to primary, returning production to the original region with minimal disruption.

Why C is wrong: A data share only provides read-only access and cannot transfer the primary role, so the original account could not resume serving writes through a share, making this unsuitable for failback.

Why D is wrong: Manual export and import is error-prone, can miss changes, and bypasses the managed replication and promotion mechanism, so it is neither the supported nor the least disruptive way to fail back.

Free sampleData Collaborationeasy

A data provider wants to publish a curated dataset so that any Snowflake customer can discover it, request access, and query it themselves, without the provider having to know each consumer's account identifier in advance. Which Snowflake feature best meets this requirement?

  • AA public listing on the Snowflake Marketplace, because publishing the dataset as a listing lets any Snowflake customer discover it in the catalogue and obtain live access on request Correct
  • BA direct secure data share created for each consumer, because the provider can add every interested account to the share so the data stays live and consumers query it without copying
  • CDatabase replication to a secondary account, because replicating the database to each consumer's region lets them browse and query the dataset as if it were their own local copy
  • DAn external stage exposing the files, because pointing consumers at a cloud storage stage lets them copy the dataset into their own account whenever they want to query it
Recognise that a public Snowflake Marketplace listing provides open discovery and live access for consumers the provider has not identified in advance. The Snowflake Marketplace is a catalogue where providers publish listings that any Snowflake customer can find and request, with access delivered through secure data sharing so the data stays live and is not copied. This open discovery is what distinguishes a Marketplace listing from a direct secure share, which must target consumer accounts the provider already knows by identifier.

Why A is correct: A public Marketplace listing places the dataset in a catalogue every Snowflake customer can browse, so consumers the provider has never met can discover it and obtain live, query-ready access without the provider naming their accounts first.

Why B is wrong: A direct secure share keeps data live and copy-free, but it must target named consumer accounts the provider already knows, so it cannot offer open discovery to customers whose account identifiers the provider has not been given.

Why C is wrong: Database replication copies a database between accounts the same organisation controls for availability, not a discovery channel for external customers, so it neither lists the data publicly nor lets unknown consumers request access.

Why D is wrong: An external stage exposes files in cloud storage and forces consumers to copy data in, which gives no catalogue discovery, no managed access requests, and no live shared dataset, so it does not fit the requirement.

Free sampleData Collaborationeasy

A provider wants to distribute an application that runs stored procedures and user-defined functions inside each consumer's own Snowflake account, so the provider's proprietary logic executes on the consumer's data without the provider seeing that data. Which Snowflake capability is designed for this?

  • AA secure data share containing the procedures, because sharing the database that holds the stored procedures lets each consumer run the provider's logic over their own data in place
  • BA scheduled task in the provider account, because a task can reach into each consumer account on a schedule to execute the provider's procedures against the consumer's data
  • CAn application built with the Snowflake Native Apps Framework, because it packages procedures and functions that install and run inside the consumer account against the consumer's own data Correct
  • DA materialized view published to the Marketplace, because the view precomputes the provider's transformations and refreshes them automatically inside each consumer account
Recognise that the Snowflake Native Apps Framework packages provider logic to install and run inside a consumer account on the consumer's own data. The Native Apps Framework lets providers package application logic, including stored procedures and user-defined functions, into a distributable app that consumers install and run within their own account. The code executes against the consumer's data locally, so the provider distributes proprietary logic and the consumer keeps their data private, which plain data sharing of read-only objects cannot achieve.

Why A is wrong: A secure data share grants read access to data objects and shared views, but it is built to share data rather than to package and install an application with its own logic running in the consumer account, so it does not fit.

Why B is wrong: A task runs only within its own account and cannot execute against another account's data, so it cannot deliver provider logic into consumer accounts and is the wrong mechanism entirely.

Why C is correct: The Snowflake Native Apps Framework lets a provider package logic such as stored procedures and functions into an app that installs and executes inside the consumer's account, so the provider's code runs on consumer data without the provider accessing it.

Why D is wrong: A materialized view stores precomputed query results, not executable application logic such as procedures and functions, so it cannot package and distribute a running application to consumers.

Other domains in this exam

See also the COF-C03 cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Snowflake. Original, blueprint-aligned practice material only.